Evaluating the Effectiveness of ARIMA and SARIMA Models for PM2.5 Forecasting in Bangladesh: A Time-Series Study (2000–2026)

Evaluating the Effectiveness of ARIMA and SARIMA Models for PM2.5 Forecasting in Bangladesh: A Time-Series Study (2000–2026) | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Evaluating the Effectiveness of ARIMA and SARIMA Models for PM2.5 Forecasting in Bangladesh: A Time-Series Study (2000–2026) Md. Mahadi Hassan This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7348600/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Background and Aim: Particulate matter ≤ 2.5 micrometers (PM2.5) is a major air pollutant linked to serious environmental and public health hazards. In Bangladesh, PM2.5 levels often exceed WHO guidelines due to unplanned urbanization, deforestation, industrial emissions, and vehicular pollution. This study explores long-term trends and seasonal variations in PM2.5 concentrations in Bangladesh and forecasts future levels using time-series models—Autoregressive Integrated Moving Average (ARIMA) and Seasonal ARIMA (SARIMA). Methods Monthly average PM2.5 data (2000–2024) were obtained from NASA’s Giovanni platform. Forecasts for 2025 and 2026 were generated using R’s auto.arima() function, which selected the best models based on Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), Mean Absolute Percentage Error (MAPE), and Mean Absolute Error (MAE). Results ARIMA(3,0,1) and SARIMA(1,0,0)(1,1,2)[ 12 ] were identified as optimal models based on statistical criteria. Both projected similar overall trends, but while the ARIMA model showed a stable trend, the SARIMA model captured seasonal fluctuations in PM2.5 levels. The Ljung-Box test confirmed SARIMA’s superior performance in accounting for white noise, highlighting the importance of seasonal components in accurate forecasting. Conclusion This study demonstrates the value of ARIMA and SARIMA models for analyzing and predicting air pollution trends in Bangladesh. These models, supported by strong statistical validation, provide effective tools for environmental monitoring and policymaking. Accurate PM2.5 forecasts can support timely interventions, inform public health strategies, and guide the development of early warning systems to reduce pollution-related health risks. Environmental Chemistry particulate matter PM2.5 air pollution ARIMA SARIMA Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 1. Introduction 1.1 Background Air pollution poses a substantial threat to the environment and human health because of its multifaceted impacts on the ecosystem, biological system, and physiological processes [ 1 ], [ 2 ], [ 3 ]. Several components residing in the atmosphere, which are responsible for air pollution, are called air pollutants [ 4 ], [ 5 ]. Particulate matter with a diameter of 2.5 µm or less are known as PM2.5 in short, is a major contributor to global air pollution [ 6 ], [ 7 ]. It poses a significant threat to human health due to its ability to penetrate deep into the respiratory tract and bloodstream, contributing to a wide range of health issues, including respiratory infections, cardiovascular diseases, and premature mortality [ 8 ], [ 9 ], [ 10 ], [ 11 ]. In Bangladesh, rapid urbanization, unregulated industrialization, and biomass burning have exacerbated PM2.5 pollution, with annual concentrations exceeding WHO guidelines by 10–15 times [ 12 ], [ 13 ], [ 14 ]. Cities like Dhaka rank among the most polluted globally, with PM2.5 levels linked to causing respiratory diseases, cardiovascular morbidity, and reduced life expectancy among the general population [ 15 ], [ 16 ]. 1.2 Problem Statement and Novelty Bangladesh consistently ranks among the countries with the worst air quality globally. Dhaka, the capital city, has repeatedly been listed among the most polluted cities, with PM2.5 concentrations exceeding WHO-recommended levels most of the time [ 17 ]. Despite growing evidence linking PM2.5 to adverse health and economic outcomes, limited research has systematically analyzed long-term PM2.5 trends using robust statistical models [ 18 ]. The absence of predictive modeling impedes the ability of policymakers and environmental agencies to implement timely interventions. Existing monitoring efforts, while improving, still lack integrated forecasting tools that can project future pollutant concentrations to guide ecological planning and public health strategies. While prior studies have explored PM2.5 pollution in Bangladesh using satellite or observational data [ 6 ], [ 19 ], [ 20 ], very few have employed advanced time-series forecasting techniques such as ARIMA and SARIMA to model and predict long-term pollutant behavior. This study is among the first to apply and compare ARIMA and SARIMA models to a 25-year dataset of PM2.5 concentrations, offering statistically validated forecasts. The incorporation of seasonality using SARIMA provides an added dimension often overlooked in earlier studies, which primarily focused on descriptive analyses or monthly averages. This methodological rigor enhances the precision of forecasting, which is critical for air quality management. 1.3 Study Significance This research holds substantial relevance for environmental policymakers, urban planners, and public health officials. Accurate forecasting of PM2.5 trends can support evidence-based decision-making to mitigate exposure risks and facilitate the development of timely interventions. In the context of Bangladesh, where air pollution is a major contributor to disease burden, the ability to anticipate pollution peaks can aid in issuing health advisories, optimizing traffic flow, regulating emissions, and improving community awareness. Furthermore, the study adds to the limited but growing body of literature on air quality forecasting in low- and middle-income countries (LMICs), providing a replicable model that can be adapted for other pollutants or regions. 1.4 Study Objectives The primary objective of this study was to analyze and forecast the long-term trends of PM2.5 concentrations in Bangladesh from 2000 to 2024 using Autoregressive Integrated Moving Average (ARIMA) and Seasonal ARIMA (SARIMA) models. Specific objectives included: Examining the historical trend of PM2.5 concentrations using time series analysis. Developing and validating ARIMA and SARIMA models to fit the observed data. Generating short-term (24 months) forecasts to identify potential future pollution patterns. Comparing the performance of ARIMA and SARIMA models in capturing the temporal dynamics of PM2.5. 2. Materials and Methods 2.1 Study Design A secondary data-based observational study design was employed for this study. The monthly average PM2.5 atmospheric concentration in Bangladesh during 2000–2024 was analyzed for insights and forecasting of the PM2.5 trend. The STROBE guideline was followed for reporting the findings of the study. 2.2 Data Source The data used in this study is publicly accessible and free to download and was obtained using Giovanni ( https://giovanni.gsfc.nasa.gov/giovanni/ ), a web application system developed by NASA for displaying and analyzing geophysical parameters in remote sensing data, where the data lineage is known. The monthly average concentration of PM2.5 was collected for Bangladesh from 2000–2024 time period. The filters applied for data selection were – Select Plot : Time-series Area Averaged, Duration : 01-01-2000 to 31-12-2024, Measurements : Particulate Matter, Select Region : Bangladesh, Temporal Resolutions : monthly. 2.3 Statistical Analysis The compatibility for ARIMA and SARIMA modelling of the data was evaluated before the model selection for the trend analysis and forecasting. Relevant analyses and visualizations were done in RStudio using R (4.4.2). The following R packages were imported and used for analyzing the data, generating the plots: dplyr, tidyr, forecast, tseries, ggplot2, and lubridate. 2.4 ARIMA Model The ARIMA (Autoregressive Integrated Moving Average) model, introduced by Box and Jenkins [ 21 ], is a widely used statistical method for time series forecasting. It consists of three key components: Autoregressive (AR) – Captures dependency on past observations, Integration (I) – Differencing renders non-stationary data stationary, and Moving Average (MA) – Accounts for error-term dependencies. The general form of an ARIMA(p, d, q) model is expressed as: $$\:\left(1-\:\sum\:_{i=1}^{p}{\varphi\:}_{i}{L}^{i}\right){\left(1-L\right)}^{d}{X}_{t}=c+\left(1+\:\sum\:_{j=1}^{q}{\theta\:}_{j}{L}^{j}\right){Z}_{t}$$ Here, X t is the differenced time series, L denotes the lag operator (shifting observations backward in time), ϕ 1 , ϕ 2 , …, ϕ p are autoregressive coefficients, d is the differencing order, θ 1 , θ 2 , …, θ q are moving average coefficients, Z t represents white noise error, and c is a constant. The model first stabilizes non-stationary data via differencing \(\:{(1-L)}^{d}{X}_{t}\) , then applies AR and MA components to the stationary series. Optimal values for p (AR order), d (differencing degree), and q (MA order) must be determined empirically for each dataset. 2.5 SARIMA Model The SARIMA (Seasonal Autoregressive Integrated Moving Average) model is an extension of the ARIMA framework designed to handle seasonal patterns in time series data, as formalized by Box and Jenkins [ 21 ]. This model incorporates three core elements: Seasonal Autoregressive (SAR) - Models dependence on seasonal lags, Integration (I) - Differencing removes trend and seasonal non-stationarity, and Seasonal Moving Average (SMA) - Accounts for seasonal error dependencies. The general formulation of a SARIMA(p,d,q)(P,D,Q)s model is given by: $$\:\left(1-\sum\:_{i=1}^{p}{\varphi\:}_{i}{L}^{i}\right)\left(1-\sum\:_{k=1}^{P}{\varPhi\:}_{k}{L}^{k.s}\right){(1-L)}^{d}{\left(1-{L}^{s}\right)}^{D}{X}_{t}=(1+\sum\:_{j=1}^{q}{\theta\:}_{j}{L}^{j})(1+\sum\:_{m=1}^{Q}{\varTheta\:}_{m}{L}^{m.s}){Z}_{t}$$ Here, X t represents the differenced time series (with both regular and seasonal differencing applied), L is the lag operator, s denotes the seasonal period (e.g., 12 for monthly data), ϕ 1 , ϕ 2 , …, ϕ p ​ are non-seasonal autoregressive coefficients, Φ 1 , Φ 2 , …, Φ p are seasonal autoregressive coefficients, d is the degree of non-seasonal differencing, D is the degree of seasonal differencing, θ 1 , θ 2 , …, θ q are non-seasonal moving average coefficients, Θ 1 , Θ 2 , …, Θ Q are seasonal moving average coefficients, and Z t is a white noise process. The model first applies regular differencing \(\:{\left(1-L\right)}^{d}\) and seasonal differencing \(\:{\left(1-{L}^{s}\right)}^{D}\) to achieve stationarity, then combines both non-seasonal and seasonal AR/MA components. The seasonal terms (with period, s) capture repeating patterns, while the non-seasonal terms model short-term dependencies. Optimal model orders (p,d,q) for non-seasonal components and (P,D,Q) for seasonal components must be identified through rigorous model selection procedures. 2.6 Model Fitting The ARIMA and SARIMA model fitting was done by following the five-step structured and systematic process described by Hyndman and Athanasopoulos for automatic model selection based on evaluation metrics [ 22 ]. The steps followed for model selection are – Plotting data for identifying trend, seasonality, and outliers. Transforming data if necessary Using automated algorithms to find and fit the best model parameters Checking the residuals selected is white noise Finalizing and evaluating the model 2.7 Evaluation Metrics 2.7.1 Akaike Information Criterion (AIC) The Akaike Information Criterion (AIC) evaluates the relative quality of statistical models by balancing goodness-of-fit against model complexity. It is widely used for model selection, with lower values indicating a superior fit. The AIC penalizes the likelihood function (L) based on the number of estimated parameters (k) [ 23 ]. $$\:\text{A}\text{I}\text{C}=\:-2\text{l}\text{n}\left(\text{L}\right)+2\text{k}$$ Here, L is the maximized likelihood of the model, and k is the number of parameters. AIC favors models that achieve high likelihood with parsimony, though it may overfit with large datasets. 2.7.2 Bayesian Information Criterion (BIC) The Bayesian Information Criterion (BIC) extends the AIC by imposing a stricter penalty for model complexity, thereby favoring simpler models [ 24 ]. It incorporates the sample size (n) to mitigate overfitting: $$\:\text{B}\text{I}\text{C}=\:-2\text{l}\text{n}\left(\text{L}\right)+\text{k}\text{l}\text{n}\left(\text{n}\right)$$ Here, L is the maximized likelihood of the model, k is the number of parameters, and n is the number of observations. BIC is particularly robust for large datasets, as its penalty term grows logarithmically with sample size. 2.7.3 Mean Absolute Percentage Error (MAPE) The Mean Absolute Percentage Error (MAPE) quantifies forecast accuracy as a percentage of the actual values, making it intuitively interpretable for relative error assessment [ 25 ]. $$\:\text{M}\text{A}\text{P}\text{E}=\:\frac{100\%}{n}\sum\:_{t=1}^{n}\frac{|{A}_{t}-{F}_{t}|}{\left|{A}_{t}\right|}$$ Here, A t is the actual value at time t, F t is the forecasted value, and n is the number of observations. 2.7.4 Mean Absolute Error (MAE) The Mean Absolute Error measures average forecast error magnitude without considering direction, providing scale-dependent accuracy assessment [ 26 ]. $$\:\text{M}\text{A}\text{E}=\:\frac{1}{n}\sum\:_{t=1}^{n}|{A}_{t}-{F}_{t}|$$ Here, A t ​ = actual value at time t, F t ​ = forecasted value, n = number of observations 2.8 Ethics Statement This study utilized remote sensing data, which involved no human or animal participants. Therefore, ethical approval from an Institutional Review Board (IRB) is not required. 3. Results 3.1 Plot and Data Transformation Figure 1 visualizes the monthly atmospheric PM2.5 concentration in Bangladesh over the period. From the figure, the seasonal fluctuations of PM2.5 are evident, indicating a sharp increase at the start of the year with a gradual decline at the year-end. The highest concentration was observed exceeding the value of 100 µg/m 3 , while the lowest was below 25 µg/m 3 during the study period. However, a more in-depth component analysis was required for the understanding of the PM2.5 data. Figure 2 shows the decomposition of the time series into its trend, seasonal, and residual components. It reveals distinct components of trend, seasonality, and random fluctuations. From the plot, a long-term upward pattern in PM2.5 levels, particularly noticeable after 2010, suggests a gradual deterioration of air quality over the years. The seasonality displays a clear and consistent periodic pattern, with higher concentrations occurring during the winter months and lower values in the monsoon and summer periods, likely due to changes in meteorological conditions and emission sources. The residual (remainder) captures short-term irregular variations not explained by the trend or seasonality. Most residuals appear to be within a reasonable range, suggesting the stationarity of the data. Overall, the decomposition supports the presence of strong seasonality, which justifies the use of ARIMA and SARIMA modeling to account for both components in forecasting future PM2.5 levels. 3.2 Data Stationarity From the autocorrelation and partial autocorrelation plot, no significant spike was observed for the original time series data. This suggests the stationarity of the data, indicating no need for differencing. However, the ACF and PACF plots are not necessary for the stationarity check. To confirm the absence of stationarity, the augmented Dickey-Fuller (ADF) test and the Box-Ljung test were conducted (ADF Statistics: -13.05, p-value = 0.001 and Box-Ljung Statistics: 396.81, p-value < 0.001), which indeed confirmed the presence of stationarity appropriate for ARIMA and SARIMA modelling. However, this was later confirmed by the auto.arima function, which eventually selected d = 0 for fitting both ARIMA and SARIMA models, while selecting the best model outperforming other parameter combinations. 3.3 Model Fitting The ARIMA and SARIMA model parameters were selected using the “auto.arima” function from the “forecast” package in R. Both ARIMA and SARIMA models can be fit using a particular data setting the argument seasonal as FALSE for ARIMA models and TRUE for SARIMA models. Based on the evaluation metrics, the function selected ARIMA (3,0,1) and SARIMA (1,0,0)(1,1,2)[ 12 ] as the best models fitted for our data. The evaluation metrics of both models are presented in Table 1 . Table 1 Performance Evaluation of the Fitted ARIMA and SARIMA Models Model AIC BIC MAE MAPE ARIMA (3,0,1) 2339.741 2361.963 8.875 25.428 SARIMA (1,0,0)(1,1,2)[ 12 ] 2136.074 2158.057 6.346 16.655 Abbreviations: AIC, Akaike’s Information Criterion; BIC, Bayesian Information Criterion; MAE, Mean Absolute Error; MAPE, Mean Absolute Percent Error 3.4 Residual Diagnostics Residual diagnostic tests were conducted to evaluate the adequacy of the fitted ARIMA and SARIMA models. In addition, the Ljung-Box test was employed to assess whether the residuals exhibited significant autocorrelation, which would indicate a poor model fit. For the ARIMA (3,0,1) model (Fig. 4 ), the Ljung-Box test produced a test statistic of Q = 101.31* with 20 degrees of freedom, resulting in a p-value < 0.001 (p = 7.34 × 10⁻¹³). This highly significant result indicates the presence of autocorrelation in the residuals, suggesting that the model did not sufficiently capture the underlying patterns in the data. Consequently, the residuals are not consistent with white noise, which undermines the reliability of the ARIMA model’s forecasts. In contrast, the SARIMA(1,0,0)(1,1,2)[ 12 ] model (Fig. 5 ) demonstrated substantially better performance in terms of residual behavior. The Ljung-Box test yielded a Q = 19.477* with 20 degrees of freedom, resulting in a non-significant p-value of 0.4911. This indicates that the residuals are independently distributed and consistent with white noise, implying that the SARIMA model adequately captured the autocorrelation structure in the data. Overall, the SARIMA model is the more appropriate choice for forecasting in this instance. Unlike the ARIMA model, the SARIMA model satisfies the assumptions of residual independence and randomness, thereby offering a more reliable basis for prediction. 3.5 Forecasting Figure 6 illustrates the forecasted monthly PM2.5 concentrations for the period January 2025 to December 2026 using both ARIMA and SARIMA models. The ARIMA model forecasts a relatively smooth and stable trend, with concentrations fluctuating narrowly between approximately 43 to 50 µg/m³. In contrast, the SARIMA model captures clear seasonal variations, with predicted peaks during the winter months (e.g., January and December) and troughs during the monsoon and summer periods (e.g., June to August). This seasonal pattern aligns with previously observed historical trends in PM2.5 concentrations in Bangladesh. Notably, the confidence intervals associated with the SARIMA forecasts are narrower than those of the ARIMA model across most months, suggesting that SARIMA provides more precise forecasts with less uncertainty. This reflects SARIMA’s advantage in modeling seasonal components explicitly, allowing it to capture both short-term fluctuations and long-term trends more effectively. Based on that, the SARIMA model seems more appropriate for forecasting PM2.5 concentrations in contexts where strong seasonal patterns exist, while ARIMA may underestimate such seasonal effects by producing smoothed projections with broader uncertainty. 4. Discussion Particulate matter has become a significant concern in the context of air pollution and public health in Bangladesh over the past decades due to its increasing emissions and consequent rise in atmospheric concentration [ 27 ]. As a critical air pollutant, PM2.5 exerts a substantial influence on both global and national air quality [ 28 ]. The rapid pace of urbanization, unplanned infrastructure development, widespread use of motor vehicles, and unregulated industrial practices have collectively contributed to the continuous release and accumulation of fine particulate matter in the atmosphere [ 29 ], [ 30 ]. In light of these issues, a comprehensive analysis of long-term trends and projections of atmospheric PM2.5 concentrations is essential to enhance understanding of the current situation, assess potential health and environmental impacts, and formulate effective strategies to mitigate the associated risks. Previous studies have explored different methodologies and models for gaining a better understanding and forecasting of the PM2.5 situation in Bangladesh. Spatiotemporal analysis conducted by Shi et al. revealed the increased level of PM2.5 in Bangladesh for the 1999 to 2014 period [ 31 ]. Their analysis also revealed that Bangladesh had the highest level of PM2.5 concentration among all the South and Southeast Asian countries. In addition, they forecasted a steady increase in the PM2.5 level in Bangladesh over time. This is similar to the findings of our study, where our trend analysis revealed that indeed the PM2.5 concentration was particularly high in 2015 and 2016 compared to previous years and was found to have consistently higher concentrations. A study conducted by Shahriar et al. [ 32 ] evaluated the effectiveness of several hybrid models for trend analysis and forecasting of PM2.5. In that study, the ARIMA-ANN and ARIMA-SVM showed a Mean Absolute Error of 11.37 and 14.03, respectively, which is significantly higher than the one in our study. This indicates a better fitting of the data and more accurate forecasting compared to the one where hybrid ARIMA models were used. This significant performance could be due to the use of a large dataset in our study compared to the one conducted using hybrid models. This suggests the importance of using a large dataset while analyzing and predicting time series data. Similarly, a study was conducted in Sylhet, Bangladesh, which used the Support Vector Machine (SVM) algorithm to predict the PM2.5 level and hotspots [ 33 ]. The prediction suggested a decrease in the PM2.5 concentration over time. This finding contrasts with the findings of our study, where both the ARIMA and SARIMA models predicted a consistency in the atmospheric PM2.5 concentration in 2025 and 2026. This difference could be due to the introduction of seasonality during model development. The findings strengthen the accuracy and effectiveness of ARIMA and SARIMA models for trend analysis and future forecasting, as they outperformed previously used methods and models in the Bangladesh context. The application of the SARIMA model for forecasting PM2.5 concentrations in Bangladesh represents a novel contribution of this study. To date, this is among the first investigations to conduct a comparative analysis using both ARIMA and SARIMA models in the context of air pollution in Bangladesh. This dual-model approach has facilitated a more comprehensive understanding of the performance and applicability of each method. SARIMA, which is well-established for its capacity to handle seasonal data with high variability, was particularly relevant given the temporal fluctuations in PM2.5 levels. Evaluating its effectiveness was essential for assessing the feasibility of employing these time-series models to support evidence-based decision-making regarding PM2.5 mitigation. According to the performance metrics, the SARIMA model outperformed several previously applied approaches, including spatial analysis techniques, machine learning algorithms, and hybrid models such as ARIMA-ANN, ARIMA-SVM, and standalone SVM, highlighting its potential as a reliable forecasting tool in the environmental health domain [ 31 ], [ 32 ], [ 33 ]. However, this study has several limitations. Firstly, both ARIMA and SARIMA are univariate time-series models that analyze and forecast trends based solely on a single variable—in this case, PM2.5 concentration—over time. These models do not account for the influence of other atmospheric components and pollutants that may significantly impact PM2.5 levels. Incorporating multivariate analyses that consider various climatic and atmospheric parameters could yield more comprehensive insights and improve the accuracy of prediction systems. Future research should explore the development and application of multivariate models that integrate additional environmental variables for more robust PM2.5 forecasting. Furthermore, this study relied on satellite-derived data, which may underestimate the actual concentrations experienced by individuals at ground level. This limitation could be addressed by incorporating data from ground-based monitoring stations into the modeling process. Combining surface-level observational data with ARIMA or SARIMA models and extending the analysis to multivariate frameworks would enhance the reliability and representativeness of the findings, making them more reflective of real-world exposure levels. 5. Conclusion This study provides a comprehensive analysis of PM2.5 trends in Bangladesh from 2000 to 2024 using ARIMA and SARIMA time-series models. Findings indicate a sustained and concerning concentration of fine particulate matter, reflecting the growing environmental and public health challenges posed by air pollution. The forecasting models offer valuable insights into the future trajectory of PM2.5 concentrations, which can serve as a critical tool for environmental planning and decision-making. In light of the projected trends, immediate and coordinated action is required to curb PM2.5 emissions. Strengthening regulatory frameworks to control industrial emissions, promoting cleaner transportation systems, and enforcing stricter vehicle emission standards are essential steps. Furthermore, the adoption of sustainable urban planning practices and the promotion of green infrastructure can significantly contribute to improving air quality. Equally important is the need for public awareness and community engagement. Educating citizens about the sources and health impacts of PM2.5 pollution can foster behavioral changes and encourage active participation in air quality improvement initiatives. Integrating air pollution topics into educational curricula and launching nationwide awareness campaigns can help build a more informed and responsible population. By aligning scientific forecasting with targeted mitigation strategies and community-driven efforts, Bangladesh can make meaningful progress toward improving air quality and protecting public health for future generations Declarations Author Contributions Md. Mahadi Hassan: Conceptualization, Methodology, Formal analysis, Visualization, Writing – Original Draft, Writing – Review & Editing. Conflicts of Interest All authors have consented to the publication of this article. Data Availability Statement The data used in this study are open and accessible at Giovanni (https://giovanni.gsfc.nasa.gov/giovanni/) and can be downloaded by applying the filters described in the methods section of the study. Funding This study received funding from no external sources at any stage of the publication process. Transparency statement The lead author, Md. Mahadi Hassan affirms that this manuscript is an honest, accurate, and transparent account of the study being reported; that no important aspects of the study have been omitted; and that any discrepancies from the study as planned (and, if relevant, registered) have been explained. Acknowledgments Special thanks to Noushin Nohor, Md. Shahidul Hassan and Marium Hassan for their valuable guidance, inspiration, and support throughout the journey. References M. Kampa and E. Castanas, “Human health effects of air pollution,” Environ. Pollut. Barking Essex 1987 , vol. 151, no. 2, pp. 362–367, Jan. 2008, doi: 10.1016/j.envpol.2007.06.012. M. A. S. Fahim and J. S. Visockienė, “Air pollution sources and their impact on the environment,” Moksl. – Liet. Ateitis Sci. – Future Lith. , vol. 16, Apr. 2024, doi: 10.3846/mla.2024.21293. H. Paramesh, D. J. Christopher, and J. S. Menon, “Air pollution, climate change and the health sector: Linkages and solutions,” in Climate Change and the Health Sector , Routledge India, 2021. L. W. Canter, “Sources, Effects, and Fate of Pollutants,” in Air Pollution , CRC Press, 2000. M. K. Theodore and L. Theodore, “Classification and Sources of Pollutants,” in Introduction to Environmental Management , 2nd ed., CRC Press, 2021. J. R. Mitra and K. Czajkowski, “Spatiotemporal patterns and hot spots of PM2.5 in Bangladesh,” Atmospheric Res. , vol. 315, p. 107898, Apr. 2025, doi: 10.1016/j.atmosres.2024.107898. O. US EPA, “Particulate Matter (PM) Basics.” Accessed: May 22, 2025. [Online]. Available: https://www.epa.gov/pm-pollution/particulate-matter-pm-basics Y.-F. Xing, Y.-H. Xu, M.-H. Shi, and Y.-X. Lian, “The impact of PM2.5 on the human respiratory system,” J. Thorac. Dis. , vol. 8, no. 1, pp. E69–E74, Jan. 2016, doi: 10.3978/j.issn.2072-1439.2016.01.19. P. Thangavel, D. Park, and Y.-C. Lee, “Recent Insights into Particulate Matter (PM2.5)-Mediated Toxicity in Humans: An Overview,” Int. J. Environ. Res. Public. Health , vol. 19, no. 12, p. 7511, Jun. 2022, doi: 10.3390/ijerph19127511. F. Chanda et al. , “PM2.5-mediated cardiovascular disease in aging: Cardiometabolic risks, molecular mechanisms and potential interventions,” Sci. Total Environ. , vol. 954, p. 176255, Dec. 2024, doi: 10.1016/j.scitotenv.2024.176255. N. R. Martins and G. Carrilho da Graça, “Impact of PM2.5 in indoor urban environments: A review,” Sustain. Cities Soc. , vol. 42, pp. 259–275, Oct. 2018, doi: 10.1016/j.scs.2018.07.011. T. Zarin and Md. Esraz-Ul-Zannat, “Assessing the potential impacts of LULC change on urban air quality in Dhaka city,” Ecol. Indic. , vol. 154, p. 110746, Oct. 2023, doi: 10.1016/j.ecolind.2023.110746. F. Ahmed, A. Z. M. Bayazid, M. Islam, M. Rahaman, and F. Al Muntasir, “The terrible air pollution in Dhaka city is getting worse,” GSC Adv. Res. Rev. , vol. 19, pp. 42–052, Apr. 2024, doi: 10.30574/gscarr.2024.19.1.0133. A.-A.- Faisal et al. , “Assessment of temporal shifting of PM2.5, lockdown effect, and influences of seasonal meteorological factors over the fastest-growing megacity, Dhaka,” Spat. Inf. Res. , vol. 30, no. 3, pp. 441–453, Jun. 2022, doi: 10.1007/s41324-022-00441-w. M. A. Rahaman, A. Kalam, and Md. Al-Mamun, “Unplanned urbanization and health risks of Dhaka City in Bangladesh: uncovering the associations between urban environment and public health,” Front. Public Health , vol. 11, p. 1269362, Oct. 2023, doi: 10.3389/fpubh.2023.1269362. R. Fuller et al. , “Pollution and health: a progress update,” Lancet Planet. Health , vol. 6, no. 6, pp. e535–e547, Jun. 2022, doi: 10.1016/S2542-5196(22)00090-0. IQAir, “2023 IQAir World Air Quality Report.” Accessed: May 23, 2025. [Online]. Available: https://www.iqair.com/newsroom/waqr-2023-pr Z. Ma et al. , “A review of statistical methods used for developing large-scale and long-term PM2.5 models from satellite data,” Remote Sens. Environ. , vol. 269, p. 112827, Feb. 2022, doi: 10.1016/j.rse.2021.112827. M. Rahman and L. Meng, “Examining the Spatial and Temporal Variation of PM2.5 and Its Linkage with Meteorological Conditions in Dhaka, Bangladesh,” Atmosphere , vol. 15, no. 12, Art. no. 12, Dec. 2024, doi: 10.3390/atmos15121426. M. S. Hassan, M. A. H. Bhuiyan, and M. T. Rahman, “Sources, pattern, and possible health impacts of PM2.5 in the central region of Bangladesh using PMF, SOM, and machine learning techniques,” Case Stud. Chem. Environ. Eng. , vol. 8, p. 100366, Dec. 2023, doi: 10.1016/j.cscee.2023.100366. G. E. P. Box and G. M. Jenkins, Time Series Analysis: Forecasting and Control . Holden-Day, 1976. R. J. Hyndman and G. Athanasopoulos, “Forecasting: Principles and Practice”. H. Bozdogan, “Model selection and Akaike’s Information Criterion (AIC): The general theory and its analytical extensions,” Psychometrika , vol. 52, no. 3, pp. 345–370, Sep. 1987, doi: 10.1007/BF02294361. G. Schwarz, “Estimating the Dimension of a Model,” Ann. Stat. , vol. 6, no. 2, pp. 461–464, Mar. 1978, doi: 10.1214/aos/1176344136. S. Kim and H. Kim, “A new metric of absolute percentage error for intermittent demand forecasts,” Int. J. Forecast. , vol. 32, no. 3, pp. 669–679, Jul. 2016, doi: 10.1016/j.ijforecast.2015.12.003. C. J. Willmott and K. Matsuura, “Advantages of the mean absolute error (MAE) over the root mean square error (RMSE) in assessing average model performance,” Clim. Res. , vol. 30, no. 1, pp. 79–82, 2005. T. B. Dibya, A. Y. Proma, and S. M. R. Dewan, “Poor Respiratory Health is a Consequence of Dhaka’s Polluted Air: A Bangladeshi Perspective,” Environ. Health Insights , vol. 17, p. 11786302231206126, Oct. 2023, doi: 10.1177/11786302231206126. S. Sangkham et al. , “An update on adverse health effects from exposure to PM2.5,” Environ. Adv. , vol. 18, p. 100603, Dec. 2024, doi: 10.1016/j.envadv.2024.100603. P. Purohit et al. , “Mitigation pathways towards national ambient air quality standards in India,” Environ. Int. , vol. 133, p. 105147, Dec. 2019, doi: 10.1016/j.envint.2019.105147. S. Wang, L. Xu, S. Ge, J. Jiao, B. Pan, and Y. Shu, “Driving force heterogeneity of urban PM2.5 pollution: Evidence from the Yangtze River Delta, China,” Ecol. Indic. , vol. 113, p. 106210, Jun. 2020, doi: 10.1016/j.ecolind.2020.106210. Y. Shi, T. Matsunaga, Y. Yamaguchi, Z. Li, X. Gu, and X. Chen, “Long-term trends and spatial patterns of satellite-retrieved PM2.5 concentrations in South and Southeast Asia from 1999 to 2014,” Sci. Total Environ. , vol. 615, pp. 177–186, Feb. 2018, doi: 10.1016/j.scitotenv.2017.09.241. S. A. Shahriar et al. , “Potential of ARIMA-ANN, ARIMA-SVM, DT and CatBoost for Atmospheric PM2.5 Forecasting in Bangladesh,” Atmosphere , vol. 12, no. 1, Art. no. 1, Jan. 2021, doi: 10.3390/atmos12010100. M. Rahman, L. Meng, A. J. Mathews, and S. Bertman, “Spatiotemporal Analysis of Urban Growth and PM2.5 Concentrations in Sylhet, Bangladesh,” Atmosphere , vol. 15, no. 11, Art. no. 11, Nov. 2024, doi: 10.3390/atmos15111305. Additional Declarations The authors declare no competing interests. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7348600","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":498950817,"identity":"a44ae985-2ae8-4f91-a67e-ccb8b0d3644d","order_by":0,"name":"Md. Mahadi Hassan","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABCElEQVRIiWNgGAWjYHCDBAZmBgZmHn4wuwCvUsYGBgYDhBbJBhDbgAQtDAYHQIJ4tPC39x5/8KPmD4Nue3bi54Iaaxnj86sTPzwwYJDnFzuAVYvEmXOJjT3HDBjMzrzdLD3jWDqP2Y23myWADjOcOTsBqxYDiRzDBt4GoJYbuRukedgOA7Wc3QDSkmBwG7eWxr8QLZt/8/w7zGM84+zmH4S0NENt2SbN23aYx4C/dxteWyTOnDGcLXPMmAfol23WvH3pPBI3eLdZJBhI4PQLf3uPwcc3NXJyZsdzN9/m+WZtz99/dvPNHxU28vzS2LXAAA+SxWCVEniVo1t8gBTVo2AUjIJRMAIAANyHXk3iQlodAAAAAElFTkSuQmCC","orcid":"https://orcid.org/0009-0004-9615-3729","institution":"Department of Public Health and Informatics, Jahangirnagar University, Dhaka, Bangladesh","correspondingAuthor":true,"prefix":"","firstName":"Md.","middleName":"Mahadi","lastName":"Hassan","suffix":""}],"badges":[],"createdAt":"2025-08-11 17:14:33","currentVersionCode":1,"declarations":{"humanSubjects":false,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":false,"humanSubjectConsent":false,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-7348600/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7348600/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":88892843,"identity":"1d46025f-64c6-4e30-bbd7-4567e64af6ae","added_by":"auto","created_at":"2025-08-12 12:54:18","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":189502,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eMonthly Average Atmospheric PM2.5 Concentration (µg/m\u003c/strong\u003e\u003csup\u003e\u003cstrong\u003e3\u003c/strong\u003e\u003c/sup\u003e\u003cstrong\u003e) in Bangladesh (2000-2024)\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7348600/v1/3de324b92a51ae599ce9a083.png"},{"id":88892838,"identity":"0074d52e-25c4-4527-957c-62bc5524297c","added_by":"auto","created_at":"2025-08-12 12:54:18","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":50825,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eDecomposition Plot of the Original PM2.5 Concentration\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7348600/v1/ecfc64fafd3fc51fd267df8a.png"},{"id":88897959,"identity":"3b29772a-84b5-407b-b8f9-1e116ce92a3d","added_by":"auto","created_at":"2025-08-12 13:18:18","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":16587,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eACF and PACF Plots of the Observed Data\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7348600/v1/e1875226f01fba900e4c6dab.png"},{"id":88892839,"identity":"61c7fdea-49e2-435c-8057-4d2b168a681b","added_by":"auto","created_at":"2025-08-12 12:54:18","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":40946,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eResidual Diagnostics for the ARIMA (3,0,1) Model\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7348600/v1/4dddf74b1827a6d21d4526a5.png"},{"id":88894407,"identity":"eccd2048-9608-403a-a9e6-787f1f2ce93c","added_by":"auto","created_at":"2025-08-12 13:02:18","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":40701,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eResidual Diagnostic for the SARIMA (1,0,0)(1,1,2)[12] Model\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"floatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-7348600/v1/d9b2a90b09f98c7b5b0b0ed9.png"},{"id":88896916,"identity":"47a62fdc-56c6-4b0b-822b-b9c64542aea7","added_by":"auto","created_at":"2025-08-12 13:10:18","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":164756,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eForecasted PM2.5 Concentration (2025-2026) for Bangladesh\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"floatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-7348600/v1/36a956824a8f084f1a464588.png"},{"id":88899216,"identity":"bd683900-7dab-4989-b135-99fad1e16a40","added_by":"auto","created_at":"2025-08-12 13:26:19","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1492205,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7348600/v1/6ec3ebff-2371-4318-b4c0-4dd22f732b15.pdf"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003e\u003cstrong\u003eEvaluating the Effectiveness of ARIMA and SARIMA Models for PM2.5 Forecasting in Bangladesh: A Time-Series Study (2000–2026)\u003c/strong\u003e\u003c/p\u003e","fulltext":[{"header":"1. Introduction","content":"\u003cdiv id=\"Sec2\" class=\"Section2\"\u003e\u003ch2\u003e1.1 Background\u003c/h2\u003e\u003cp\u003eAir pollution poses a substantial threat to the environment and human health because of its multifaceted impacts on the ecosystem, biological system, and physiological processes [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e], [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e], [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. Several components residing in the atmosphere, which are responsible for air pollution, are called air pollutants [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e], [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]. Particulate matter with a diameter of 2.5 \u0026micro;m or less are known as PM2.5 in short, is a major contributor to global air pollution [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e], [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. It poses a significant threat to human health due to its ability to penetrate deep into the respiratory tract and bloodstream, contributing to a wide range of health issues, including respiratory infections, cardiovascular diseases, and premature mortality [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e], [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e], [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e], [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. In Bangladesh, rapid urbanization, unregulated industrialization, and biomass burning have exacerbated PM2.5 pollution, with annual concentrations exceeding WHO guidelines by 10\u0026ndash;15 times [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e], [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e], [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]. Cities like Dhaka rank among the most polluted globally, with PM2.5 levels linked to causing respiratory diseases, cardiovascular morbidity, and reduced life expectancy among the general population [\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e], [\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e].\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003e1.2 Problem Statement and Novelty\u003c/h2\u003e\u003cp\u003eBangladesh consistently ranks among the countries with the worst air quality globally. Dhaka, the capital city, has repeatedly been listed among the most polluted cities, with PM2.5 concentrations exceeding WHO-recommended levels most of the time [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]. Despite growing evidence linking PM2.5 to adverse health and economic outcomes, limited research has systematically analyzed long-term PM2.5 trends using robust statistical models [\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]. The absence of predictive modeling impedes the ability of policymakers and environmental agencies to implement timely interventions. Existing monitoring efforts, while improving, still lack integrated forecasting tools that can project future pollutant concentrations to guide ecological planning and public health strategies. While prior studies have explored PM2.5 pollution in Bangladesh using satellite or observational data [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e], [\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e], [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e], very few have employed advanced time-series forecasting techniques such as ARIMA and SARIMA to model and predict long-term pollutant behavior. This study is among the first to apply and compare ARIMA and SARIMA models to a 25-year dataset of PM2.5 concentrations, offering statistically validated forecasts. The incorporation of seasonality using SARIMA provides an added dimension often overlooked in earlier studies, which primarily focused on descriptive analyses or monthly averages. This methodological rigor enhances the precision of forecasting, which is critical for air quality management.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec4\" class=\"Section2\"\u003e\u003ch2\u003e1.3 Study Significance\u003c/h2\u003e\u003cp\u003eThis research holds substantial relevance for environmental policymakers, urban planners, and public health officials. Accurate forecasting of PM2.5 trends can support evidence-based decision-making to mitigate exposure risks and facilitate the development of timely interventions. In the context of Bangladesh, where air pollution is a major contributor to disease burden, the ability to anticipate pollution peaks can aid in issuing health advisories, optimizing traffic flow, regulating emissions, and improving community awareness. Furthermore, the study adds to the limited but growing body of literature on air quality forecasting in low- and middle-income countries (LMICs), providing a replicable model that can be adapted for other pollutants or regions.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec5\" class=\"Section2\"\u003e\u003ch2\u003e1.4 Study Objectives\u003c/h2\u003e\u003cp\u003eThe primary objective of this study was to analyze and forecast the long-term trends of PM2.5 concentrations in Bangladesh from 2000 to 2024 using Autoregressive Integrated Moving Average (ARIMA) and Seasonal ARIMA (SARIMA) models. Specific objectives included:\u003c/p\u003e\u003cp\u003e\u003col\u003e\u003cspan\u003e\u003cli\u003e\u003cp\u003eExamining the historical trend of PM2.5 concentrations using time series analysis.\u003c/p\u003e\u003c/li\u003e\u003c/span\u003e\u003cspan\u003e\u003cli\u003e\u003cp\u003eDeveloping and validating ARIMA and SARIMA models to fit the observed data.\u003c/p\u003e\u003c/li\u003e\u003c/span\u003e\u003cspan\u003e\u003cli\u003e\u003cp\u003eGenerating short-term (24 months) forecasts to identify potential future pollution patterns.\u003c/p\u003e\u003c/li\u003e\u003c/span\u003e\u003cspan\u003e\u003cli\u003e\u003cp\u003eComparing the performance of ARIMA and SARIMA models in capturing the temporal dynamics of PM2.5.\u003c/p\u003e\u003c/li\u003e\u003c/span\u003e\u003c/ol\u003e\u003c/p\u003e\u003c/div\u003e"},{"header":"2. Materials and Methods","content":"\u003cdiv id=\"Sec7\" class=\"Section2\"\u003e\u003ch2\u003e2.1 Study Design\u003c/h2\u003e\u003cp\u003eA secondary data-based observational study design was employed for this study. The monthly average PM2.5 atmospheric concentration in Bangladesh during 2000\u0026ndash;2024 was analyzed for insights and forecasting of the PM2.5 trend. The STROBE guideline was followed for reporting the findings of the study.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\u003ch2\u003e2.2 Data Source\u003c/h2\u003e\u003cp\u003eThe data used in this study is publicly accessible and free to download and was obtained using Giovanni (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://giovanni.gsfc.nasa.gov/giovanni/\u003c/span\u003e\u003cspan address=\"https://giovanni.gsfc.nasa.gov/giovanni/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e), a web application system developed by NASA for displaying and analyzing geophysical parameters in remote sensing data, where the data lineage is known. The monthly average concentration of PM2.5 was collected for Bangladesh from 2000\u0026ndash;2024 time period. The filters applied for data selection were \u0026ndash; \u003cb\u003eSelect Plot\u003c/b\u003e: Time-series Area Averaged, \u003cb\u003eDuration\u003c/b\u003e: 01-01-2000 to 31-12-2024, \u003cb\u003eMeasurements\u003c/b\u003e: Particulate Matter, \u003cb\u003eSelect Region\u003c/b\u003e: Bangladesh, \u003cb\u003eTemporal Resolutions\u003c/b\u003e: monthly.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e\u003ch2\u003e2.3 Statistical Analysis\u003c/h2\u003e\u003cp\u003eThe compatibility for ARIMA and SARIMA modelling of the data was evaluated before the model selection for the trend analysis and forecasting. Relevant analyses and visualizations were done in RStudio using R (4.4.2). The following R packages were imported and used for analyzing the data, generating the plots: dplyr, tidyr, forecast, tseries, ggplot2, and lubridate.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e\u003ch2\u003e2.4 ARIMA Model\u003c/h2\u003e\u003cp\u003eThe ARIMA (Autoregressive Integrated Moving Average) model, introduced by Box and Jenkins [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e], is a widely used statistical method for time series forecasting. It consists of three key components: Autoregressive (AR) \u0026ndash; Captures dependency on past observations, Integration (I) \u0026ndash; Differencing renders non-stationary data stationary, and Moving Average (MA) \u0026ndash; Accounts for error-term dependencies. The general form of an ARIMA(p, d, q) model is expressed as:\u003cdiv id=\"Equa\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equa\" name=\"EquationSource\"\u003e\n$$\\:\\left(1-\\:\\sum\\:_{i=1}^{p}{\\varphi\\:}_{i}{L}^{i}\\right){\\left(1-L\\right)}^{d}{X}_{t}=c+\\left(1+\\:\\sum\\:_{j=1}^{q}{\\theta\\:}_{j}{L}^{j}\\right){Z}_{t}$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eHere, X\u003csub\u003et\u003c/sub\u003e is the differenced time series, \u003cem\u003eL\u003c/em\u003e denotes the lag operator (shifting observations backward in time), ϕ\u003csub\u003e1\u003c/sub\u003e, ϕ\u003csub\u003e2\u003c/sub\u003e, \u0026hellip;, ϕ\u003csub\u003ep\u003c/sub\u003e are autoregressive coefficients, d is the differencing order, θ\u003csub\u003e1\u003c/sub\u003e, θ\u003csub\u003e2\u003c/sub\u003e, \u0026hellip;, θ\u003csub\u003eq\u003c/sub\u003e are moving average coefficients, Z\u003csub\u003et\u003c/sub\u003e represents white noise error, and c is a constant.\u003c/p\u003e\u003cp\u003eThe model first stabilizes non-stationary data via differencing \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{(1-L)}^{d}{X}_{t}\\)\u003c/span\u003e\u003c/span\u003e, then applies AR and MA components to the stationary series. Optimal values for p (AR order), d (differencing degree), and q (MA order) must be determined empirically for each dataset.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\u003ch2\u003e2.5 SARIMA Model\u003c/h2\u003e\u003cp\u003eThe SARIMA (Seasonal Autoregressive Integrated Moving Average) model is an extension of the ARIMA framework designed to handle seasonal patterns in time series data, as formalized by Box and Jenkins [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]. This model incorporates three core elements: Seasonal Autoregressive (SAR) - Models dependence on seasonal lags, Integration (I) - Differencing removes trend and seasonal non-stationarity, and Seasonal Moving Average (SMA) - Accounts for seasonal error dependencies. The general formulation of a SARIMA(p,d,q)(P,D,Q)s model is given by:\u003cdiv id=\"Equb\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equb\" name=\"EquationSource\"\u003e\n$$\\:\\left(1-\\sum\\:_{i=1}^{p}{\\varphi\\:}_{i}{L}^{i}\\right)\\left(1-\\sum\\:_{k=1}^{P}{\\varPhi\\:}_{k}{L}^{k.s}\\right){(1-L)}^{d}{\\left(1-{L}^{s}\\right)}^{D}{X}_{t}=(1+\\sum\\:_{j=1}^{q}{\\theta\\:}_{j}{L}^{j})(1+\\sum\\:_{m=1}^{Q}{\\varTheta\\:}_{m}{L}^{m.s}){Z}_{t}$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eHere, X\u003csub\u003et\u003c/sub\u003e represents the differenced time series (with both regular and seasonal differencing applied), L is the lag operator, s denotes the seasonal period (e.g., 12 for monthly data), ϕ\u003csub\u003e1\u003c/sub\u003e, ϕ\u003csub\u003e2\u003c/sub\u003e, \u0026hellip;, ϕ\u003csub\u003ep\u003c/sub\u003e​ are non-seasonal autoregressive coefficients, Φ\u003csub\u003e1\u003c/sub\u003e, Φ\u003csub\u003e2\u003c/sub\u003e, \u0026hellip;, Φ\u003csub\u003ep\u003c/sub\u003e are seasonal autoregressive coefficients, d is the degree of non-seasonal differencing, D is the degree of seasonal differencing, θ\u003csub\u003e1\u003c/sub\u003e, θ\u003csub\u003e2\u003c/sub\u003e, \u0026hellip;, θ\u003csub\u003eq\u003c/sub\u003e are non-seasonal moving average coefficients, Θ\u003csub\u003e1\u003c/sub\u003e, Θ\u003csub\u003e2\u003c/sub\u003e, \u0026hellip;, Θ\u003csub\u003eQ\u003c/sub\u003e are seasonal moving average coefficients, and Z\u003csub\u003et\u003c/sub\u003e is a white noise process.\u003c/p\u003e\u003cp\u003eThe model first applies regular differencing \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\left(1-L\\right)}^{d}\\)\u003c/span\u003e\u003c/span\u003e and seasonal differencing \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\left(1-{L}^{s}\\right)}^{D}\\)\u003c/span\u003e\u003c/span\u003e to achieve stationarity, then combines both non-seasonal and seasonal AR/MA components. The seasonal terms (with period, s) capture repeating patterns, while the non-seasonal terms model short-term dependencies. Optimal model orders (p,d,q) for non-seasonal components and (P,D,Q) for seasonal components must be identified through rigorous model selection procedures.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\u003ch2\u003e2.6 Model Fitting\u003c/h2\u003e\u003cp\u003eThe ARIMA and SARIMA model fitting was done by following the five-step structured and systematic process described by Hyndman and Athanasopoulos for automatic model selection based on evaluation metrics [\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]. The steps followed for model selection are \u0026ndash;\u003c/p\u003e\u003cp\u003e\u003col\u003e\u003cspan\u003e\u003cli\u003e\u003cp\u003ePlotting data for identifying trend, seasonality, and outliers.\u003c/p\u003e\u003c/li\u003e\u003c/span\u003e\u003cspan\u003e\u003cli\u003e\u003cp\u003eTransforming data if necessary\u003c/p\u003e\u003c/li\u003e\u003c/span\u003e\u003cspan\u003e\u003cli\u003e\u003cp\u003eUsing automated algorithms to find and fit the best model parameters\u003c/p\u003e\u003c/li\u003e\u003c/span\u003e\u003cspan\u003e\u003cli\u003e\u003cp\u003eChecking the residuals selected is white noise\u003c/p\u003e\u003c/li\u003e\u003c/span\u003e\u003cspan\u003e\u003cli\u003e\u003cp\u003eFinalizing and evaluating the model\u003c/p\u003e\u003c/li\u003e\u003c/span\u003e\u003c/ol\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\u003ch2\u003e2.7 Evaluation Metrics\u003c/h2\u003e\u003cdiv id=\"Sec14\" class=\"Section3\"\u003e\u003ch2\u003e2.7.1 Akaike Information Criterion (AIC)\u003c/h2\u003e\u003cp\u003eThe Akaike Information Criterion (AIC) evaluates the relative quality of statistical models by balancing goodness-of-fit against model complexity. It is widely used for model selection, with lower values indicating a superior fit. The AIC penalizes the likelihood function (L) based on the number of estimated parameters (k) [\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e].\u003cdiv id=\"Equc\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equc\" name=\"EquationSource\"\u003e\n$$\\:\\text{A}\\text{I}\\text{C}=\\:-2\\text{l}\\text{n}\\left(\\text{L}\\right)+2\\text{k}$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eHere, L is the maximized likelihood of the model, and k is the number of parameters. AIC favors models that achieve high likelihood with parsimony, though it may overfit with large datasets.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec15\" class=\"Section3\"\u003e\u003ch2\u003e2.7.2 Bayesian Information Criterion (BIC)\u003c/h2\u003e\u003cp\u003eThe Bayesian Information Criterion (BIC) extends the AIC by imposing a stricter penalty for model complexity, thereby favoring simpler models [\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e]. It incorporates the sample size (n) to mitigate overfitting:\u003cdiv id=\"Equd\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equd\" name=\"EquationSource\"\u003e\n$$\\:\\text{B}\\text{I}\\text{C}=\\:-2\\text{l}\\text{n}\\left(\\text{L}\\right)+\\text{k}\\text{l}\\text{n}\\left(\\text{n}\\right)$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eHere, L is the maximized likelihood of the model, k is the number of parameters, and n is the number of observations. BIC is particularly robust for large datasets, as its penalty term grows logarithmically with sample size.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec16\" class=\"Section3\"\u003e\u003ch2\u003e2.7.3 Mean Absolute Percentage Error (MAPE)\u003c/h2\u003e\u003cp\u003eThe Mean Absolute Percentage Error (MAPE) quantifies forecast accuracy as a percentage of the actual values, making it intuitively interpretable for relative error assessment [\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e].\u003cdiv id=\"Eque\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Eque\" name=\"EquationSource\"\u003e\n$$\\:\\text{M}\\text{A}\\text{P}\\text{E}=\\:\\frac{100\\%}{n}\\sum\\:_{t=1}^{n}\\frac{|{A}_{t}-{F}_{t}|}{\\left|{A}_{t}\\right|}$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eHere, A\u003csub\u003et\u003c/sub\u003e is the actual value at time t, F\u003csub\u003et\u003c/sub\u003e is the forecasted value, and n is the number of observations.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec17\" class=\"Section3\"\u003e\u003ch2\u003e2.7.4 Mean Absolute Error (MAE)\u003c/h2\u003e\u003cp\u003eThe Mean Absolute Error measures average forecast error magnitude without considering direction, providing scale-dependent accuracy assessment [\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e].\u003cdiv id=\"Equf\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equf\" name=\"EquationSource\"\u003e\n$$\\:\\text{M}\\text{A}\\text{E}=\\:\\frac{1}{n}\\sum\\:_{t=1}^{n}|{A}_{t}-{F}_{t}|$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eHere, A\u003csub\u003et\u003c/sub\u003e​ = actual value at time t, F\u003csub\u003et\u003c/sub\u003e​ = forecasted value, n\u0026thinsp;=\u0026thinsp;number of observations\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv id=\"Sec18\" class=\"Section2\"\u003e\u003ch2\u003e2.8 Ethics Statement\u003c/h2\u003e\u003cp\u003eThis study utilized remote sensing data, which involved no human or animal participants. Therefore, ethical approval from an Institutional Review Board (IRB) is not required.\u003c/p\u003e\u003c/div\u003e"},{"header":"3. Results","content":"\u003cdiv id=\"Sec20\" class=\"Section2\"\u003e\u003ch2\u003e3.1 Plot and Data Transformation\u003c/h2\u003e\u003cp\u003eFigure \u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e visualizes the monthly atmospheric PM2.5 concentration in Bangladesh over the period. From the figure, the seasonal fluctuations of PM2.5 are evident, indicating a sharp increase at the start of the year with a gradual decline at the year-end. The highest concentration was observed exceeding the value of 100 \u0026micro;g/m\u003csup\u003e3\u003c/sup\u003e, while the lowest was below 25 \u0026micro;g/m\u003csup\u003e3\u003c/sup\u003e during the study period.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eHowever, a more in-depth component analysis was required for the understanding of the PM2.5 data. Figure\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e shows the decomposition of the time series into its trend, seasonal, and residual components. It reveals distinct components of trend, seasonality, and random fluctuations. From the plot, a long-term upward pattern in PM2.5 levels, particularly noticeable after 2010, suggests a gradual deterioration of air quality over the years. The seasonality displays a clear and consistent periodic pattern, with higher concentrations occurring during the winter months and lower values in the monsoon and summer periods, likely due to changes in meteorological conditions and emission sources. The residual (remainder) captures short-term irregular variations not explained by the trend or seasonality. Most residuals appear to be within a reasonable range, suggesting the stationarity of the data. Overall, the decomposition supports the presence of strong seasonality, which justifies the use of ARIMA and SARIMA modeling to account for both components in forecasting future PM2.5 levels.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec21\" class=\"Section2\"\u003e\u003ch2\u003e3.2 Data Stationarity\u003c/h2\u003e\u003cp\u003eFrom the autocorrelation and partial autocorrelation plot, no significant spike was observed for the original time series data. This suggests the stationarity of the data, indicating no need for differencing. However, the ACF and PACF plots are not necessary for the stationarity check. To confirm the absence of stationarity, the augmented Dickey-Fuller (ADF) test and the Box-Ljung test were conducted (ADF Statistics: -13.05, \u003cem\u003ep-value\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.001 and Box-Ljung Statistics: 396.81, \u003cem\u003ep-value\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001), which indeed confirmed the presence of stationarity appropriate for ARIMA and SARIMA modelling. However, this was later confirmed by the auto.arima function, which eventually selected d\u0026thinsp;=\u0026thinsp;0 for fitting both ARIMA and SARIMA models, while selecting the best model outperforming other parameter combinations.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec22\" class=\"Section2\"\u003e\u003ch2\u003e3.3 Model Fitting\u003c/h2\u003e\u003cp\u003eThe ARIMA and SARIMA model parameters were selected using the \u0026ldquo;auto.arima\u0026rdquo; function from the \u0026ldquo;forecast\u0026rdquo; package in R. Both ARIMA and SARIMA models can be fit using a particular data setting the argument seasonal as FALSE for ARIMA models and TRUE for SARIMA models. Based on the evaluation metrics, the function selected ARIMA (3,0,1) and SARIMA (1,0,0)(1,1,2)[\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e] as the best models fitted for our data. The evaluation metrics of both models are presented in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003ePerformance Evaluation of the Fitted ARIMA and SARIMA Models\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"5\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eModel\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eAIC\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eBIC\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eMAE\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eMAPE\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eARIMA (3,0,1)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e2339.741\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e2361.963\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e8.875\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e25.428\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSARIMA (1,0,0)(1,1,2)[\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e2136.074\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e2158.057\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e6.346\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e16.655\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colspan=\"5\" nameend=\"c5\" namest=\"c1\"\u003e\u003cp\u003eAbbreviations: AIC, Akaike\u0026rsquo;s Information Criterion; BIC, Bayesian Information Criterion; MAE, Mean Absolute Error; MAPE, Mean Absolute Percent Error\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec23\" class=\"Section2\"\u003e\u003ch2\u003e3.4 Residual Diagnostics\u003c/h2\u003e\u003cp\u003eResidual diagnostic tests were conducted to evaluate the adequacy of the fitted ARIMA and SARIMA models. In addition, the Ljung-Box test was employed to assess whether the residuals exhibited significant autocorrelation, which would indicate a poor model fit.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eFor the ARIMA (3,0,1) model (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e), the Ljung-Box test produced a test statistic of Q\u0026thinsp;=\u0026thinsp;101.31* with 20 degrees of freedom, resulting in a p-value\u0026thinsp;\u0026lt;\u0026thinsp;0.001 (p\u0026thinsp;=\u0026thinsp;7.34 \u0026times; 10⁻\u0026sup1;\u0026sup3;). This highly significant result indicates the presence of autocorrelation in the residuals, suggesting that the model did not sufficiently capture the underlying patterns in the data. Consequently, the residuals are not consistent with white noise, which undermines the reliability of the ARIMA model\u0026rsquo;s forecasts.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eIn contrast, the SARIMA(1,0,0)(1,1,2)[\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e] model (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e) demonstrated substantially better performance in terms of residual behavior. The Ljung-Box test yielded a Q\u0026thinsp;=\u0026thinsp;19.477* with 20 degrees of freedom, resulting in a non-significant p-value of 0.4911. This indicates that the residuals are independently distributed and consistent with white noise, implying that the SARIMA model adequately captured the autocorrelation structure in the data.\u003c/p\u003e\u003cp\u003eOverall, the SARIMA model is the more appropriate choice for forecasting in this instance. Unlike the ARIMA model, the SARIMA model satisfies the assumptions of residual independence and randomness, thereby offering a more reliable basis for prediction.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec24\" class=\"Section2\"\u003e\u003ch2\u003e3.5 Forecasting\u003c/h2\u003e\u003cp\u003eFigure \u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e illustrates the forecasted monthly PM2.5 concentrations for the period January 2025 to December 2026 using both ARIMA and SARIMA models. The ARIMA model forecasts a relatively smooth and stable trend, with concentrations fluctuating narrowly between approximately 43 to 50 \u0026micro;g/m\u0026sup3;. In contrast, the SARIMA model captures clear seasonal variations, with predicted peaks during the winter months (e.g., January and December) and troughs during the monsoon and summer periods (e.g., June to August). This seasonal pattern aligns with previously observed historical trends in PM2.5 concentrations in Bangladesh. Notably, the confidence intervals associated with the SARIMA forecasts are narrower than those of the ARIMA model across most months, suggesting that SARIMA provides more precise forecasts with less uncertainty. This reflects SARIMA\u0026rsquo;s advantage in modeling seasonal components explicitly, allowing it to capture both short-term fluctuations and long-term trends more effectively. Based on that, the SARIMA model seems more appropriate for forecasting PM2.5 concentrations in contexts where strong seasonal patterns exist, while ARIMA may underestimate such seasonal effects by producing smoothed projections with broader uncertainty.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e"},{"header":"4. Discussion","content":"\u003cp\u003eParticulate matter has become a significant concern in the context of air pollution and public health in Bangladesh over the past decades due to its increasing emissions and consequent rise in atmospheric concentration [\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e]. As a critical air pollutant, PM2.5 exerts a substantial influence on both global and national air quality [\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e]. The rapid pace of urbanization, unplanned infrastructure development, widespread use of motor vehicles, and unregulated industrial practices have collectively contributed to the continuous release and accumulation of fine particulate matter in the atmosphere [\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e], [\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e]. In light of these issues, a comprehensive analysis of long-term trends and projections of atmospheric PM2.5 concentrations is essential to enhance understanding of the current situation, assess potential health and environmental impacts, and formulate effective strategies to mitigate the associated risks.\u003c/p\u003e\u003cp\u003ePrevious studies have explored different methodologies and models for gaining a better understanding and forecasting of the PM2.5 situation in Bangladesh. Spatiotemporal analysis conducted by Shi et al. revealed the increased level of PM2.5 in Bangladesh for the 1999 to 2014 period [\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e]. Their analysis also revealed that Bangladesh had the highest level of PM2.5 concentration among all the South and Southeast Asian countries. In addition, they forecasted a steady increase in the PM2.5 level in Bangladesh over time. This is similar to the findings of our study, where our trend analysis revealed that indeed the PM2.5 concentration was particularly high in 2015 and 2016 compared to previous years and was found to have consistently higher concentrations. A study conducted by Shahriar et al. [\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e] evaluated the effectiveness of several hybrid models for trend analysis and forecasting of PM2.5. In that study, the ARIMA-ANN and ARIMA-SVM showed a Mean Absolute Error of 11.37 and 14.03, respectively, which is significantly higher than the one in our study. This indicates a better fitting of the data and more accurate forecasting compared to the one where hybrid ARIMA models were used. This significant performance could be due to the use of a large dataset in our study compared to the one conducted using hybrid models. This suggests the importance of using a large dataset while analyzing and predicting time series data. Similarly, a study was conducted in Sylhet, Bangladesh, which used the Support Vector Machine (SVM) algorithm to predict the PM2.5 level and hotspots [\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e]. The prediction suggested a decrease in the PM2.5 concentration over time. This finding contrasts with the findings of our study, where both the ARIMA and SARIMA models predicted a consistency in the atmospheric PM2.5 concentration in 2025 and 2026. This difference could be due to the introduction of seasonality during model development. The findings strengthen the accuracy and effectiveness of ARIMA and SARIMA models for trend analysis and future forecasting, as they outperformed previously used methods and models in the Bangladesh context.\u003c/p\u003e\u003cp\u003eThe application of the SARIMA model for forecasting PM2.5 concentrations in Bangladesh represents a novel contribution of this study. To date, this is among the first investigations to conduct a comparative analysis using both ARIMA and SARIMA models in the context of air pollution in Bangladesh. This dual-model approach has facilitated a more comprehensive understanding of the performance and applicability of each method. SARIMA, which is well-established for its capacity to handle seasonal data with high variability, was particularly relevant given the temporal fluctuations in PM2.5 levels. Evaluating its effectiveness was essential for assessing the feasibility of employing these time-series models to support evidence-based decision-making regarding PM2.5 mitigation. According to the performance metrics, the SARIMA model outperformed several previously applied approaches, including spatial analysis techniques, machine learning algorithms, and hybrid models such as ARIMA-ANN, ARIMA-SVM, and standalone SVM, highlighting its potential as a reliable forecasting tool in the environmental health domain [\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e], [\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e], [\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eHowever, this study has several limitations. Firstly, both ARIMA and SARIMA are univariate time-series models that analyze and forecast trends based solely on a single variable\u0026mdash;in this case, PM2.5 concentration\u0026mdash;over time. These models do not account for the influence of other atmospheric components and pollutants that may significantly impact PM2.5 levels. Incorporating multivariate analyses that consider various climatic and atmospheric parameters could yield more comprehensive insights and improve the accuracy of prediction systems. Future research should explore the development and application of multivariate models that integrate additional environmental variables for more robust PM2.5 forecasting. Furthermore, this study relied on satellite-derived data, which may underestimate the actual concentrations experienced by individuals at ground level. This limitation could be addressed by incorporating data from ground-based monitoring stations into the modeling process. Combining surface-level observational data with ARIMA or SARIMA models and extending the analysis to multivariate frameworks would enhance the reliability and representativeness of the findings, making them more reflective of real-world exposure levels.\u003c/p\u003e"},{"header":"5. Conclusion","content":"\u003cp\u003eThis study provides a comprehensive analysis of PM2.5 trends in Bangladesh from 2000 to 2024 using ARIMA and SARIMA time-series models. Findings indicate a sustained and concerning concentration of fine particulate matter, reflecting the growing environmental and public health challenges posed by air pollution. The forecasting models offer valuable insights into the future trajectory of PM2.5 concentrations, which can serve as a critical tool for environmental planning and decision-making. In light of the projected trends, immediate and coordinated action is required to curb PM2.5 emissions. Strengthening regulatory frameworks to control industrial emissions, promoting cleaner transportation systems, and enforcing stricter vehicle emission standards are essential steps. Furthermore, the adoption of sustainable urban planning practices and the promotion of green infrastructure can significantly contribute to improving air quality.\u003c/p\u003e\u003cp\u003eEqually important is the need for public awareness and community engagement. Educating citizens about the sources and health impacts of PM2.5 pollution can foster behavioral changes and encourage active participation in air quality improvement initiatives. Integrating air pollution topics into educational curricula and launching nationwide awareness campaigns can help build a more informed and responsible population. By aligning scientific forecasting with targeted mitigation strategies and community-driven efforts, Bangladesh can make meaningful progress toward improving air quality and protecting public health for future generations\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAuthor Contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMd. Mahadi Hassan:\u003c/strong\u003e Conceptualization, Methodology, Formal analysis, Visualization, Writing \u0026ndash; Original Draft, Writing \u0026ndash; Review \u0026amp; Editing.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConflicts of Interest\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll authors have consented to the publication of this article.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData Availability Statement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe data used in this study are open and accessible at Giovanni (https://giovanni.gsfc.nasa.gov/giovanni/) and can be downloaded by applying the filters described in the methods section of the study.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study received funding from no external sources at any stage of the publication process.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTransparency statement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe lead author, Md. Mahadi Hassan affirms that this manuscript is an honest, accurate, and transparent account of the study being reported; that no important aspects of the study have been omitted; and that any discrepancies from the study as planned (and, if relevant, registered) have been explained.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgments\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eSpecial thanks to Noushin Nohor, Md. Shahidul Hassan and Marium Hassan for their valuable guidance, inspiration, and support throughout the journey.\u003cbr\u003e\u0026nbsp;\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eM. Kampa and E. Castanas, \u0026ldquo;Human health effects of air pollution,\u0026rdquo; \u003cem\u003eEnviron. Pollut. Barking Essex 1987\u003c/em\u003e, vol. 151, no. 2, pp. 362\u0026ndash;367, Jan. 2008, doi: 10.1016/j.envpol.2007.06.012.\u003c/li\u003e\n\u003cli\u003eM. A. S. Fahim and J. S. Visockienė, \u0026ldquo;Air pollution sources and their impact on the environment,\u0026rdquo; \u003cem\u003eMoksl. \u0026ndash; Liet. Ateitis Sci. \u0026ndash; Future Lith.\u003c/em\u003e, vol. 16, Apr. 2024, doi: 10.3846/mla.2024.21293.\u003c/li\u003e\n\u003cli\u003eH. Paramesh, D. J. Christopher, and J. S. Menon, \u0026ldquo;Air pollution, climate change and the health sector: Linkages and solutions,\u0026rdquo; in \u003cem\u003eClimate Change and the Health Sector\u003c/em\u003e, Routledge India, 2021.\u003c/li\u003e\n\u003cli\u003eL. W. Canter, \u0026ldquo;Sources, Effects, and Fate of Pollutants,\u0026rdquo; in \u003cem\u003eAir Pollution\u003c/em\u003e, CRC Press, 2000.\u003c/li\u003e\n\u003cli\u003eM. K. Theodore and L. Theodore, \u0026ldquo;Classification and Sources of Pollutants,\u0026rdquo; in \u003cem\u003eIntroduction to Environmental Management\u003c/em\u003e, 2nd ed., CRC Press, 2021.\u003c/li\u003e\n\u003cli\u003eJ. R. Mitra and K. Czajkowski, \u0026ldquo;Spatiotemporal patterns and hot spots of PM2.5 in Bangladesh,\u0026rdquo; \u003cem\u003eAtmospheric Res.\u003c/em\u003e, vol. 315, p. 107898, Apr. 2025, doi: 10.1016/j.atmosres.2024.107898.\u003c/li\u003e\n\u003cli\u003eO. US EPA, \u0026ldquo;Particulate Matter (PM) Basics.\u0026rdquo; Accessed: May 22, 2025. [Online]. Available: https://www.epa.gov/pm-pollution/particulate-matter-pm-basics\u003c/li\u003e\n\u003cli\u003eY.-F. Xing, Y.-H. Xu, M.-H. Shi, and Y.-X. Lian, \u0026ldquo;The impact of PM2.5 on the human respiratory system,\u0026rdquo; \u003cem\u003eJ. Thorac. Dis.\u003c/em\u003e, vol. 8, no. 1, pp. E69\u0026ndash;E74, Jan. 2016, doi: 10.3978/j.issn.2072-1439.2016.01.19.\u003c/li\u003e\n\u003cli\u003eP. Thangavel, D. Park, and Y.-C. Lee, \u0026ldquo;Recent Insights into Particulate Matter (PM2.5)-Mediated Toxicity in Humans: An Overview,\u0026rdquo; \u003cem\u003eInt. J. Environ. Res. Public. Health\u003c/em\u003e, vol. 19, no. 12, p. 7511, Jun. 2022, doi: 10.3390/ijerph19127511.\u003c/li\u003e\n\u003cli\u003eF. Chanda \u003cem\u003eet al.\u003c/em\u003e, \u0026ldquo;PM2.5-mediated cardiovascular disease in aging: Cardiometabolic risks, molecular mechanisms and potential interventions,\u0026rdquo; \u003cem\u003eSci. Total Environ.\u003c/em\u003e, vol. 954, p. 176255, Dec. 2024, doi: 10.1016/j.scitotenv.2024.176255.\u003c/li\u003e\n\u003cli\u003eN. R. Martins and G. Carrilho da Gra\u0026ccedil;a, \u0026ldquo;Impact of PM2.5 in indoor urban environments: A review,\u0026rdquo; \u003cem\u003eSustain. Cities Soc.\u003c/em\u003e, vol. 42, pp. 259\u0026ndash;275, Oct. 2018, doi: 10.1016/j.scs.2018.07.011.\u003c/li\u003e\n\u003cli\u003eT. Zarin and Md. Esraz-Ul-Zannat, \u0026ldquo;Assessing the potential impacts of LULC change on urban air quality in Dhaka city,\u0026rdquo; \u003cem\u003eEcol. Indic.\u003c/em\u003e, vol. 154, p. 110746, Oct. 2023, doi: 10.1016/j.ecolind.2023.110746.\u003c/li\u003e\n\u003cli\u003eF. Ahmed, A. Z. M. Bayazid, M. Islam, M. Rahaman, and F. Al Muntasir, \u0026ldquo;The terrible air pollution in Dhaka city is getting worse,\u0026rdquo; \u003cem\u003eGSC Adv. Res. Rev.\u003c/em\u003e, vol. 19, pp. 42\u0026ndash;052, Apr. 2024, doi: 10.30574/gscarr.2024.19.1.0133.\u003c/li\u003e\n\u003cli\u003eA.-A.- Faisal \u003cem\u003eet al.\u003c/em\u003e, \u0026ldquo;Assessment of temporal shifting of PM2.5, lockdown effect, and influences of seasonal meteorological factors over the fastest-growing megacity, Dhaka,\u0026rdquo; \u003cem\u003eSpat. Inf. Res.\u003c/em\u003e, vol. 30, no. 3, pp. 441\u0026ndash;453, Jun. 2022, doi: 10.1007/s41324-022-00441-w.\u003c/li\u003e\n\u003cli\u003eM. A. Rahaman, A. Kalam, and Md. Al-Mamun, \u0026ldquo;Unplanned urbanization and health risks of Dhaka City in Bangladesh: uncovering the associations between urban environment and public health,\u0026rdquo; \u003cem\u003eFront. Public Health\u003c/em\u003e, vol. 11, p. 1269362, Oct. 2023, doi: 10.3389/fpubh.2023.1269362.\u003c/li\u003e\n\u003cli\u003eR. Fuller \u003cem\u003eet al.\u003c/em\u003e, \u0026ldquo;Pollution and health: a progress update,\u0026rdquo; \u003cem\u003eLancet Planet. Health\u003c/em\u003e, vol. 6, no. 6, pp. e535\u0026ndash;e547, Jun. 2022, doi: 10.1016/S2542-5196(22)00090-0.\u003c/li\u003e\n\u003cli\u003eIQAir, \u0026ldquo;2023 IQAir World Air Quality Report.\u0026rdquo; Accessed: May 23, 2025. [Online]. Available: https://www.iqair.com/newsroom/waqr-2023-pr\u003c/li\u003e\n\u003cli\u003eZ. Ma \u003cem\u003eet al.\u003c/em\u003e, \u0026ldquo;A review of statistical methods used for developing large-scale and long-term PM2.5 models from satellite data,\u0026rdquo; \u003cem\u003eRemote Sens. Environ.\u003c/em\u003e, vol. 269, p. 112827, Feb. 2022, doi: 10.1016/j.rse.2021.112827.\u003c/li\u003e\n\u003cli\u003eM. Rahman and L. Meng, \u0026ldquo;Examining the Spatial and Temporal Variation of PM2.5 and Its Linkage with Meteorological Conditions in Dhaka, Bangladesh,\u0026rdquo; \u003cem\u003eAtmosphere\u003c/em\u003e, vol. 15, no. 12, Art. no. 12, Dec. 2024, doi: 10.3390/atmos15121426.\u003c/li\u003e\n\u003cli\u003eM. S. Hassan, M. A. H. Bhuiyan, and M. T. Rahman, \u0026ldquo;Sources, pattern, and possible health impacts of PM2.5 in the central region of Bangladesh using PMF, SOM, and machine learning techniques,\u0026rdquo; \u003cem\u003eCase Stud. Chem. Environ. Eng.\u003c/em\u003e, vol. 8, p. 100366, Dec. 2023, doi: 10.1016/j.cscee.2023.100366.\u003c/li\u003e\n\u003cli\u003eG. E. P. Box and G. M. Jenkins, \u003cem\u003eTime Series Analysis: Forecasting and Control\u003c/em\u003e. Holden-Day, 1976.\u003c/li\u003e\n\u003cli\u003eR. J. Hyndman and G. Athanasopoulos, \u0026ldquo;Forecasting: Principles and Practice\u0026rdquo;.\u003c/li\u003e\n\u003cli\u003eH. Bozdogan, \u0026ldquo;Model selection and Akaike\u0026rsquo;s Information Criterion (AIC): The general theory and its analytical extensions,\u0026rdquo; \u003cem\u003ePsychometrika\u003c/em\u003e, vol. 52, no. 3, pp. 345\u0026ndash;370, Sep. 1987, doi: 10.1007/BF02294361.\u003c/li\u003e\n\u003cli\u003eG. Schwarz, \u0026ldquo;Estimating the Dimension of a Model,\u0026rdquo; \u003cem\u003eAnn. Stat.\u003c/em\u003e, vol. 6, no. 2, pp. 461\u0026ndash;464, Mar. 1978, doi: 10.1214/aos/1176344136.\u003c/li\u003e\n\u003cli\u003eS. Kim and H. Kim, \u0026ldquo;A new metric of absolute percentage error for intermittent demand forecasts,\u0026rdquo; \u003cem\u003eInt. J. Forecast.\u003c/em\u003e, vol. 32, no. 3, pp. 669\u0026ndash;679, Jul. 2016, doi: 10.1016/j.ijforecast.2015.12.003.\u003c/li\u003e\n\u003cli\u003eC. J. Willmott and K. Matsuura, \u0026ldquo;Advantages of the mean absolute error (MAE) over the root mean square error (RMSE) in assessing average model performance,\u0026rdquo; \u003cem\u003eClim. Res.\u003c/em\u003e, vol. 30, no. 1, pp. 79\u0026ndash;82, 2005.\u003c/li\u003e\n\u003cli\u003eT. B. Dibya, A. Y. Proma, and S. M. R. Dewan, \u0026ldquo;Poor Respiratory Health is a Consequence of Dhaka\u0026rsquo;s Polluted Air: A Bangladeshi Perspective,\u0026rdquo; \u003cem\u003eEnviron. Health Insights\u003c/em\u003e, vol. 17, p. 11786302231206126, Oct. 2023, doi: 10.1177/11786302231206126.\u003c/li\u003e\n\u003cli\u003eS. Sangkham \u003cem\u003eet al.\u003c/em\u003e, \u0026ldquo;An update on adverse health effects from exposure to PM2.5,\u0026rdquo; \u003cem\u003eEnviron. Adv.\u003c/em\u003e, vol. 18, p. 100603, Dec. 2024, doi: 10.1016/j.envadv.2024.100603.\u003c/li\u003e\n\u003cli\u003eP. Purohit \u003cem\u003eet al.\u003c/em\u003e, \u0026ldquo;Mitigation pathways towards national ambient air quality standards in India,\u0026rdquo; \u003cem\u003eEnviron. Int.\u003c/em\u003e, vol. 133, p. 105147, Dec. 2019, doi: 10.1016/j.envint.2019.105147.\u003c/li\u003e\n\u003cli\u003eS. Wang, L. Xu, S. Ge, J. Jiao, B. Pan, and Y. Shu, \u0026ldquo;Driving force heterogeneity of urban PM2.5 pollution: Evidence from the Yangtze River Delta, China,\u0026rdquo; \u003cem\u003eEcol. Indic.\u003c/em\u003e, vol. 113, p. 106210, Jun. 2020, doi: 10.1016/j.ecolind.2020.106210.\u003c/li\u003e\n\u003cli\u003eY. Shi, T. Matsunaga, Y. Yamaguchi, Z. Li, X. Gu, and X. Chen, \u0026ldquo;Long-term trends and spatial patterns of satellite-retrieved PM2.5 concentrations in South and Southeast Asia from 1999 to 2014,\u0026rdquo; \u003cem\u003eSci. Total Environ.\u003c/em\u003e, vol. 615, pp. 177\u0026ndash;186, Feb. 2018, doi: 10.1016/j.scitotenv.2017.09.241.\u003c/li\u003e\n\u003cli\u003eS. A. Shahriar \u003cem\u003eet al.\u003c/em\u003e, \u0026ldquo;Potential of ARIMA-ANN, ARIMA-SVM, DT and CatBoost for Atmospheric PM2.5 Forecasting in Bangladesh,\u0026rdquo; \u003cem\u003eAtmosphere\u003c/em\u003e, vol. 12, no. 1, Art. no. 1, Jan. 2021, doi: 10.3390/atmos12010100.\u003c/li\u003e\n\u003cli\u003eM. Rahman, L. Meng, A. J. Mathews, and S. Bertman, \u0026ldquo;Spatiotemporal Analysis of Urban Growth and PM2.5 Concentrations in Sylhet, Bangladesh,\u0026rdquo; \u003cem\u003eAtmosphere\u003c/em\u003e, vol. 15, no. 11, Art. no. 11, Nov. 2024, doi: 10.3390/atmos15111305.\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"particulate matter, PM2.5, air pollution, ARIMA, SARIMA","lastPublishedDoi":"10.21203/rs.3.rs-7348600/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7348600/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground and Aim:\u003c/h2\u003e\u003cp\u003eParticulate matter\u0026thinsp;\u0026le;\u0026thinsp;2.5 micrometers (PM2.5) is a major air pollutant linked to serious environmental and public health hazards. In Bangladesh, PM2.5 levels often exceed WHO guidelines due to unplanned urbanization, deforestation, industrial emissions, and vehicular pollution. This study explores long-term trends and seasonal variations in PM2.5 concentrations in Bangladesh and forecasts future levels using time-series models\u0026mdash;Autoregressive Integrated Moving Average (ARIMA) and Seasonal ARIMA (SARIMA).\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e\u003cp\u003eMonthly average PM2.5 data (2000\u0026ndash;2024) were obtained from NASA\u0026rsquo;s Giovanni platform. Forecasts for 2025 and 2026 were generated using R\u0026rsquo;s auto.arima() function, which selected the best models based on Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), Mean Absolute Percentage Error (MAPE), and Mean Absolute Error (MAE).\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e\u003cp\u003eARIMA(3,0,1) and SARIMA(1,0,0)(1,1,2)[\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e] were identified as optimal models based on statistical criteria. Both projected similar overall trends, but while the ARIMA model showed a stable trend, the SARIMA model captured seasonal fluctuations in PM2.5 levels. The Ljung-Box test confirmed SARIMA\u0026rsquo;s superior performance in accounting for white noise, highlighting the importance of seasonal components in accurate forecasting.\u003c/p\u003e\u003ch2\u003eConclusion\u003c/h2\u003e\u003cp\u003eThis study demonstrates the value of ARIMA and SARIMA models for analyzing and predicting air pollution trends in Bangladesh. These models, supported by strong statistical validation, provide effective tools for environmental monitoring and policymaking. Accurate PM2.5 forecasts can support timely interventions, inform public health strategies, and guide the development of early warning systems to reduce pollution-related health risks.\u003c/p\u003e","manuscriptTitle":"Evaluating the Effectiveness of ARIMA and SARIMA Models for PM2.5 Forecasting in Bangladesh: A Time-Series Study (2000–2026)","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-08-12 12:54:13","doi":"10.21203/rs.3.rs-7348600/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"f8b7cb77-6be9-4ac6-bb1e-1af5b50a09fb","owner":[],"postedDate":"August 12th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":52996019,"name":"Environmental Chemistry"}],"tags":[],"updatedAt":"2025-08-12T12:54:13+00:00","versionOfRecord":[],"versionCreatedAt":"2025-08-12 12:54:13","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7348600","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7348600","identity":"rs-7348600","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}